Data from: The increased effect of spring leaf unfolding on autumn senescence in the northern and southern hemispheres
Abstract
This dataset was generated to analyze the impacts of vegetation growth carryover (VGC) and climate factors on plant phenology at a global scale. It includes satellite NDVI data from GIMMS NDVI 3g (1982–2015) and MODIS NDVI (2001–2022), as well as ground-based phenological observations from PEP 725 (1963–2015). Additionally, the dataset incorporates climate data from ERA5, including temperature, precipitation, radiation, and potential evapotranspiration, spanning 1982–2022. The dataset provides start-of-season (SOS) and end-of-season (EOS) metrics alongside these climatic variables. Covering the Northern and Southern Hemispheres, the dataset offers insights into global phenological trends, carbon cycling, and vegetation-climate interactions under changing climatic conditions. This dataset is intended for use in phenology studies, ecosystem modeling, and global change research.
https://doi.org/10.5061/dryad.7d7wm3856
Description of the data and file structure
Data and Code for Phenology and Climate Factor Analysis
This dataset and accompanying code provide the methodology and results for calculating various indices within the first 15-year moving window (1982-1996 years). The same approach can be applied to other windows for consistency.
NO1_MK_slope: The trends of phenology and climate factors are calculated over the 15-year window.
NO2_Sensitivity_Calculation: The sensitivity of phenological changes to climate factors within the window is assessed.
NO3_Contribution_calculation: Example provided for the contribution of temperature to phenological changes.
NO4_Contribution_regression: Contribution values within the window are calculated using a non-linear multivariate regression approach.
Files and variables
File: Data.rar
Description: This dataset and accompanying code provide the methodology and results for calculating various indices within the first 15-year moving window(1982-1996 years). The same approach can be applied to other windows for consistency.
Code/software
All scripts were executed using MATLAB R2020a.
NO1_MK_slope: The trends of phenology and climate factors are calculated over the 15-year window.
NO2_Sensitivity_Calculation: The sensitivity of phenological changes to climate factors within the window is assessed.
NO3_Contribution_calculation: Example provided for the contribution of temperature to phenological changes.
NO4_Contribution_regression: Contribution values within the window are calculated using a non-linear multivariate regression approach.
Access information
Other publicly accessible locations of the data:
- This dataset is only available on Dryad and has no other publicly accessible locations.
Data was derived from the following sources:
- This dataset was derived from the following publicly accessible sources:
- GIMMS NDVI 3g: Global Vegetation Greenness (NDVI) from AVHRR GIMMS-3G+, 1981-2022 | NASA Earthdata/
- MODIS NDVI: https://lpdaac.usgs.gov/products/mod13c1v006/
- PEP 725 phenological observations: http://www.pep725.eu/
- ERA5 climate data: https://confluence.ecmwf.int/display/CKB/ERA5
- Land cover dataset: https://lpdaac.usgs.gov/products/mcd12c1v006
All sources are openly accessible, and their respective usage complies with their licenses.
This dataset was collected from multiple sources. Satellite NDVI data were obtained from GIMMS NDVI 3g (1982–2015) and MODIS NDVI (2001–2022), providing global vegetation indices. Ground-based phenological observations were retrieved from the PEP 725 database (1963–2015), covering European regions. Climate data, including temperature, precipitation, radiation, and potential evapotranspiration, were extracted from ERA5 reanalysis datasets (1982–2022). Raw NDVI data were smoothed using a Savitzky-Golay filter to reduce noise. Start-of-season (SOS) and end-of-season (EOS) metrics were extracted by fitting a double logistic function to the smoothed NDVI time series. Climate data were spatially interpolated to match NDVI data resolution (0.1° × 0.1°), and all datasets were aligned temporally and spatially. Data processing and analysis were conducted in Python, with geospatial operations performed in ArcGIS.